Figaro is an end-to-end AI pipeline that takes a recorded sales call, scores it against your playbook, rewrites the rep's dialogue to show best-practice execution, and generates a coaching video of the rep's own digital twin delivering the ideal version. Fully automated. No manager time required.
Every coaching video Figaro produces follows the same automated sequence. No manager reviews. No scheduling. No 'let's circle back in our next 1:1.' The call ends. The coaching arrives.
Your rep hangs up. Figaro pulls the recording, video, and transcript from HubSpot automatically. No uploads. No tagging. No one has to remember to do anything.
Claude analyses the full transcript against your custom scoring rubric. Not a generic talk-ratio dashboard. Your rubric. Your criteria. Your definition of what a good call looks like.
The AI rewrites only the rep's side of the conversation. Every word the prospect said stays identical. The rewritten version demonstrates what the rep should have said at each moment.
Using the rep's own face and voice, Figaro generates a video of their digital twin delivering the rewritten script. Lip-synced. Natural pacing. Their actual appearance.
FFmpeg composites the AI avatar alongside the original prospect footage in a gallery-view layout that looks like the original Zoom call.
The coaching video and scorecard land in the rep's Slack DM within 30 minutes of the call ending. The team training channel gets it too.
Figaro writes the coaching video link back to the HubSpot call record. Every call now has a 'what right looks like' companion attached.
Every sales team has different criteria for what makes a good call. Figaro does not impose a generic framework. You define the rubric. You set the weights. You decide what matters.
The default rubric scores nine dimensions: discovery quality, strategy synthesis, value framing, package presentation, recommended stack confidence, cadence rationale, case study accuracy, commercial handling, and cancel policy adherence. Strategy synthesis carries the highest weight by design — it is the skill that separates closers from order-takers.
Each dimension is scored on a four-tier scale: Missing, Partial, Solid, or Exceptional. The scorecard includes direct quotes from the transcript as evidence for every score. Not a number. A verdict with proof.
Nine Dimensions
Discovery, strategy synthesis, value framing, package presentation, stack confidence, cadence rationale, case study accuracy, commercial handling, cancel policy.
Four Tiers
Missing / Partial / Solid / Exceptional. Every score backed by a direct quote from the transcript.
Fully Customisable
Replace the default rubric with your own. Add dimensions. Change weights. Make it reflect your playbook, not ours.
Every job in the system with live status tracking. Queued, scoring, generating avatar, compositing, delivering, completed, or failed. Status updates in real time without refreshing.
Side-by-side transcript comparison: original on the left, ideal on the right. Changed sections highlighted. Scorecard rendered below with direct quotes as evidence. Video player for the coaching recreation.
Avatar status, voice clone status, consent status, and performance metrics for every rep. One click to configure a new rep's digital twin.
Paste any transcript and trigger a pipeline run on demand. Use it for testing, for calls that came through a different system, or for historical call reviews.
Pulls call recordings, transcripts, and contact data. Writes coaching video links back to the call record.
Delivers the coaching video and scorecard to the rep's DM and the team training channel within 30 minutes.
Generates the avatar video using the rep's digital twin — lip-synced to the rewritten script.
Clones the rep's voice for natural-sounding audio narration in the coaching video.
Scores the transcript against your rubric and rewrites the rep's dialogue to demonstrate best-practice execution.
Composites the avatar and prospect footage into a gallery-view coaching video that matches the format of the original call.
No rep's likeness is used without explicit written consent. Figaro will not process calls for a rep who has not consented. This is enforced at the pipeline level, not just policy.
HeyGen requires every rep to complete a consent video confirmation before their avatar can be created. This is a platform-level requirement, not optional.
ElevenLabs requires the rep to submit a voice sample and verify their identity before a voice clone can be created. The clone is locked to that identity.
Your call recordings, transcripts, and coaching videos are never shared with third parties for training purposes. Data stays within your account boundaries.
Any rep can withdraw consent at any time. Their avatar and voice clone are deleted within 48 hours. Future calls are scored in text-only mode until re-consent.